Exercise for depression.
Level 1 - systematic review of randomized trials
Systematic review and meta-analysis of randomized controlled trials
PubMed 41500513 · doi:10.1002/14651858.CD004366.pub7
What was done
This Cochrane systematic review updated searches through November 2023 to evaluate exercise for depression in adults aged 18 and older. Randomised controlled trials comparing exercise against no treatment, inactive controls, or active treatments (psychological or pharmacological) were included; trials of postnatal depression were excluded. The primary outcome was depressive symptoms measured at the end of treatment and long-term follow-up using standardized mean differences (SMD). Risk of bias was evaluated using the Cochrane RoB 1 tool.
What was found
The review included 73 trials (at least 4,985 participants), with 69 contributing to meta-analyses. Compared to control interventions, exercise reduced depressive symptoms at treatment end (57 trials, n = 2,189; SMD -0.67, 95% CI -0.82 to -0.52; low certainty), though the effect attenuated in the 7 trials meeting low risk of bias criteria (n = 447; SMD -0.46, 95% CI -0.88 to -0.04). Long-term effects vs control were uncertain (9 trials, n = 405; SMD -0.53, 95% CI -1.11 to 0.06; very low certainty). Exercise showed little to no difference compared to psychological therapy (10 trials, n = 414; SMD 0.03, 95% CI -0.16 to 0.23; moderate certainty) or pharmacological therapy (5 trials, n = 330; SMD -0.11, 95% CI -0.33 to 0.10; low certainty). Treatment acceptability did not differ between groups. Adverse events in exercise groups primarily involved musculoskeletal injuries.
Why it matters
This update confirms that exercise provides moderate symptom relief for depression comparable to standard active therapies. It reinforces exercise as a viable clinical alternative or adjunct, while highlighting that the estimated effect size shrinks when restricted to higher-quality trials.
Limits
Methodological quality across the evidence base was poor: all trials carried a high risk of performance bias due to unblinded participants and providers, only 22 adequately concealed allocation, only 31 conducted intention-to-treat analyses, and 23 blinded outcome assessors. Many trials relied on self-reported mood scales. Comparisons against active treatments (psychological and pharmacological) were limited by small sample sizes and few studies, and long-term follow-up data were sparse.